Classical Mechanics, Lecture 14

February 26, 2008

lecture by John Baez
notes by Alex Hoffnung

1 The Lie Algebra of a Lie Group

We’ve seen some Lie groups already, like SO(n). Now it’s time for a formal definition:
Definition 1 A Lie group is a group G that is also a manifold such that multiplication
m: G × G → G
(g, h) 7→ gh
and inverse map
inv: G → G
g 7→ g −1
are smooth maps.
Every Lie group has a Lie algebra. To get our hands on this, given a Lie group G we define a
vector space
g = T1 G
where 1 ∈ G is the identity. We want to make this into a Lie algebra and construct an ‘exponentia-
tion’ map
exp: g → G.
But we need to start slowly, and build up some basic tools first.
Recall that given a smooth map between manifolds ψ: X → N we get a linear map called pushing
forward along ψ:
ψ∗ = dψx : Tx X → Tpsi(x) N
picture of pushforward of tangent vector
We can define ψ∗ by saying it sends [γ] ∈ Tx X where γ: R → X with γ(0) = x, to [ψγ] ∈ Tψ(x) N .

Check:
(1X )∗ = 1Tx X
and if ψ: M → N and ψ: N → P then
φ∗ ψ∗ = (φψ)∗
This idea lets us think of the g as a space of vector fields called ‘left-invariant’ vector fields:
Theorem 2 g is isomorphic to the vector space of left-invariant vector fields on G, i.e. vector
fields v ∈ V ect(G) such that
(Lg )∗ v(h) = v(gh), ∀g, h ∈ G
where left multiplication by g is:
Lg : G → G
h 7→ gh.
The isomorphism goes as follows:
{left-invariant vector fields} → g
v 7→ v(1)

1
 Before we prove this, let’s draw a picture to explain it!

picture of U (1) in the complex plane with tangent space at identity before and after left multiplication

A left-invariant vector field on U (1) is one of constant length, always pointing the same way. Given
x ∈ T1 G we get a left-invariant vector field V on G by

v(g) = (Lg )∗ x

Now let’s prove the theorem:

Proof - We construct an inverse map:

g → {left-invariant vector fields}

x 7→ v x
where
v x (h) = (Lh )∗ x ∈ Th G.
Let’s check that v x is left-invariant:

(Lg )∗ v x (h) = (Lg )∗ (Lh )∗ x

= (Lg Lh )∗ x
= (Lgh )∗ x
= v x (gh).

Next check it is an inverse map. First: start with a left-invariant w, turn it into w(1) ∈ g, then turn
that back into a left-invariant vector field v w(1) . Check: w = v w(1) .

v w(1) (h) = (Lh )∗ w(1)

= w(h).

Second: start with x ∈ g, turn it into a left-invariant vector field v x , then turn that back into
v x (1) ∈ g. Check: x = v x (1).

v x (1) = (L1 )∗ x
= (1G )∗ x
= x.

picture of Lie-algebra element going to left-invariant vector field on the circle and vice versa
We henceforth use this isomorphism to freely think of g either as T1 G or as the space of all left-
invariant vector fields on G. We use this to define a bracket operation on g, using the fact that
V ect(G) is a Lie algebra. Using g ⊆ V ect(G) to make g into a Lie algebra we just need:
Lemma 3 If v, w ∈ V ect(G) are left-invariant, so is [v, w].
Proof - For this we will use a general fact: if φ: M → N is a diffeomorphism, then given v ∈ V ect(M )

picture of push-forward of vector

If φ is a diffeomorphism there is a unique x ∈ M mapping to any y ∈ N (namely φ−1 (y)), so we can

define a vector field φ∗ v ∈ V ect(N ) by:

(φ∗ v)(y) = φ∗ v(φ−1 (y))

2
In fact, if φ: M → N is a diffeomorphism and v, w ∈ V ect(M ) then:

φ∗ [v, w] = [φ∗ v, φ∗ w].

In particular, if v and w are left-invariant vector fields on G,

(Lg )∗ v = v

(Lg )∗ w = w
so

(Lg )∗ [v, w] = [(Lg )∗ v, (Lg )∗ w]

= [v, w]

so [v, w] is left-invariant.

Now that g is a Lie algebra let’s define the exponential map

exp: g → G.

We will use a sort of hard fact

Theorem 4 If G is a Lie group, every left-invariant v ∈ V ect(G) is integrable.
(By the way: If M is a compact manifold, then every v ∈ V ect(M ) is integrable.)

Given this, any v ∈ g thought of as a left-invariant vector field, generates a flow:

φ: R × G → G

(t, g) 7→ φt (g)
and we define:
exp(tv) = φt (1) ∈ G
Example: G = U(1)
picture of circle with left-invariant vector field
tv tv
This generates a flow where φt : U (1) → U (1) is rotation by the angle i ∈ R. Note: rotation by i
is the same as multiplication by etv ∈ U (1). So

exp(tv) = φt (1) = etv 1 = etv ∈ U (1)

Whew!

Example: G = SO(n), rotation group of Rn .

This is a group of matrices, with matrix multiplication as the group operation. So SO(n) sits in the
vector space of n × n matrices:

artist’s depiction of SO(n) with its Lie algebra so(n)

In this case the Lie algebra

so(n) = {A: Rn → Rn : A linear and A∗ = −A}

and
exp: so(n) → so(n)

3
is given by
∞
X Ak
exp(A) =
k!
k=0

With this definition of exp: g → G we can check:

exp(0) = 1 ∈ G

(easy, since a flow has φ0 = 1) and:

exp((s + t)v) = exp(sv)exp(tv)

(trickier, but we use φs+t = φs φt and some more).

2 Actions of Lie Groups

What are Lie groups good for? They ‘act’ on manifolds!

Definition 5 If G is a Lie group and X is some manifold, an action of G on X is a smooth map

φ: G × X → X

(g, x) 7→ φ(g)x
such that:
φ(gh)x = φ(g)φ(h)x, ∀g, h ∈ G, ∀x ∈ X
and:
φ(1)x = x.
(These imply: φ(g −1 ) = φ(g)−1 ).

Example: A flow is the same as an action of G = R.

Example: The Euclidean group E(n) is a Lie group and it acts on Rn (“space”).

Example: The Galilei group G(n + 1) is a Lie group and it acts on Rn+1 (“spacetime”). We
also saw how G(n + 1) acts on the phase space of a free particle in Rn , X = Rn × Rn 3 (q, p).

We have seen how symmetries in G(n + 1) are related to conserved quantities, certain functions
on the Poisson manifold X. We have seen that (almost) any single function on a Poisson manifold
generates a flow, i.e. an action of R. (The vector field might fail to be integrable.)

When does a collection of functions on a Poisson manifold X give rise to an action of a Lie group G
on X? Or conversely, which group actions on X give rise to a bunch of functions on X? These are
important questions that we’ll begin to tackle next time.